ChatGPT Ads: Common Mistakes Advertisers Are Making

Since self-serve access opened to any eligible business on May 5, 2026, ChatGPT Ads has gone from an invite-only pilot with a $200,000-plus minimum to a channel almost any advertiser can test. That accessibility is exactly why mistakes are multiplying — teams that never had to think about OpenAI's ad rules, its pricing dynamics, or category restrictions are now launching campaigns fast, often without checking the parts of this channel that don't work like Google or Meta.

Here are the mistakes costing advertisers the most right now, and what to do instead.

1. Launching Without Checking Category Eligibility First

This is the mistake that stops a campaign before it ever gets a fair test.

• Adult content, gambling, alcohol, tobacco, illegal drugs, cryptocurrency, credit repair/debt settlement, and political advertising are prohibited outright — no amount of creative polish gets these approved.
• Financial services and healthcare are opening gradually through case-by-case manual review, not blanket category approval. A financial advertiser assuming their entire vertical is cleared because "finance ads are allowed now" can still get individual products rejected — credit repair, debt settlement, and alternative investments like bullion remain excluded even as other financial products open up.
• Legal services remain largely restricted, with only an exception for general legal education content — a mistake we see often is legal advertisers assuming a broad opening that hasn't actually happened.
• Confirm eligibility for your exact product, not just your general industry, before building creative or committing budget.

2. Confusing Advertiser Eligibility With Ad Placement

Even approved advertisers get this wrong. OpenAI runs two separate systems: whether your business is allowed to buy ads at all, and whether a specific ad is contextually appropriate for the specific conversation a user is having right now.

• Getting approved as an advertiser doesn't guarantee delivery into every relevant conversation — OpenAI evaluates placement conversation by conversation, and sensitive or regulated topics can be excluded from ad delivery even for fully approved advertisers.
• Don't interpret low delivery volume as a targeting problem alone; it may be a placement-eligibility issue tied to the kinds of conversations your category is allowed to appear in.
• Build this distinction into how you diagnose underperformance — the fix for a placement restriction is different from the fix for weak context hints.

3. Bidding Off Launch-Day Pricing Instead of Current Benchmarks

CPMs fell roughly 50% in the first ten weeks after launch as OpenAI scaled inventory, and CPC bidding has since become the platform's primary performance objective. Advertisers still anchoring bids to February's $60 flat CPM are working from outdated numbers.

• Current CPMs typically run $25–$60 depending on category; CPCs run roughly $3–$18, with wide variance by vertical — ecommerce at the low end, SaaS and finance considerably higher.
• Bidding too low doesn't just underperform — there's a real delivery floor. Bids under roughly $3 CPC frequently fail to win any impressions at all, regardless of how well-targeted the context hints are.
• Re-check category benchmarks regularly rather than setting a bid once at campaign launch and leaving it, since pricing has continued to move throughout 2026 as more advertisers and inventory come online.

4. Treating FTC Disclosure as the Platform's Job, Not the Advertiser's

OpenAI labels ads as "Sponsored" and visually separates them from the AI's answer — but that platform-level labeling is a baseline, not a substitute for an advertiser's own disclosure obligations.

• FTC truth-in-advertising rules require claims to be substantiated and disclosures to be clear and conspicuous, placed at the point of the claim — a footer-only disclaimer on a landing page doesn't meet that standard, and this specific failure shows up repeatedly in AI-assistant creative.
• The FTC's "Operation AI Comply" initiative has made AI-related claims a sustained enforcement priority since 2024, with continued actions into 2026. Overstating what your product does, or how "AI-powered" it is, carries real regulatory risk, not just a brand-perception risk.
• If you're running testimonials or endorsements in ChatGPT ad creative, they're subject to the FTC's Endorsement Guides — material connections between endorsers and the advertiser must be disclosed, and testimonials can't be false or misleading.

5. Ignoring Vertical-Specific Compliance Requirements

General FTC rules are only the floor. Regulated industries carry additional obligations that generic "AI advertising" guides often skip entirely.

• Financial services advertisers need to account for FINRA Rule 2210, which governs communications with the public — AI-generated ad content used in external communications counts as a communication subject to supervision and retention, not an exception because it was AI-generated.
• The CFPB's interpretive guidance treats digital marketing firms providing targeting or optimization services to financial advertisers as potentially liable under UDAAP authority — the "time or space" exception that shields some traditional media doesn't shield modern AI-targeted advertising.
• Healthcare advertisers need substantiation for any claim tied to an approved product, and unsafe, unapproved, or experimental treatment claims remain prohibited even for advertisers otherwise cleared to run healthcare ads.
• Treat vertical compliance as part of campaign build, with a disclosure and substantiation checklist reviewed before each new creative variation goes live, not a one-time legal sign-off at account setup.

6. Comparing Raw Cost to Google and Meta Without Adjusting for Intent

A common mistake in early budget conversations is treating ChatGPT Ads CPMs as simply "too expensive" next to Meta's $14–$20 CPM or Google's $5.26–$5.42 average CPC, without accounting for what the impression or click actually represents.

• ChatGPT sits at a real premium — closer to LinkedIn's roughly $39 CPM than to Meta on a raw impression basis — but the underlying argument for the channel is intent quality: a user mid-conversation, actively comparing options, is a fundamentally different impression than a user mid-scroll.
• Judging the channel purely on CPC or CPM without tracking downstream conversion rate misses the actual comparison that matters for budget decisions.
• Where budgets allow, run ChatGPT Ads as a complement funded from a dedicated test-and-learn allocation rather than forcing a head-to-head cost comparison against channels with fundamentally different audience states.

7. Skipping the Technology and Agency Partner Options That Already Fit Your Stack

Some advertisers build a fully manual, standalone ChatGPT Ads workflow when their existing media stack already has a faster path in.

• If your team already buys media through Adobe, Criteo, Kargo, Pacvue, or StackAdapt, those platforms have native API integration with ChatGPT Ads — building a separate manual process duplicates work that's already available through tools you're paying for.
• Regulated advertisers in particular often underestimate how much compliance infrastructure agency partners like Dentsu, Omnicom, Publicis, or WPP bring — disclosure review workflows and recordkeeping processes that are expensive to build in-house from scratch for a single new channel.
• This isn't a mistake of omission for every advertiser — lean teams may genuinely be better served managing directly through ads.openai.com — but it's worth an explicit evaluation rather than defaulting to fully manual management by habit.

Bringing It Together

Most of the costly ChatGPT Ads mistakes advertisers are making right now trace back to speed outpacing diligence: launching before confirming category eligibility, bidding off outdated pricing, and treating platform-level ad labeling as sufficient compliance cover. None of these mistakes are exotic — they're the kind of checklist items that get skipped when a channel opens up fast and budget follows the hype before the rulebook gets read closely. Advertisers who check eligibility, current pricing, and disclosure requirements before launch are the ones who get a fair read on whether this channel actually works for them.

Frequently Asked Questions

Why would an approved ChatGPT Ads advertiser still not see their ads delivering? Advertiser eligibility and ad placement are separate systems. Being approved to advertise doesn't guarantee delivery into every relevant conversation — OpenAI evaluates each conversation separately for contextual and topical appropriateness, which can limit delivery even for fully approved advertisers.

Are financial services and healthcare fully open for ChatGPT Ads? No. Both are opening gradually through case-by-case manual review, not blanket category approval. Specific excluded products — credit repair, debt settlement, alternative investments, unsupported medical claims — remain restricted even as other products in the same industry get approved.

Does OpenAI's "Sponsored" label satisfy FTC disclosure requirements? The platform label is a baseline, not a substitute for the advertiser's own disclosure obligations. Claims still need to be substantiated, and disclosures still need to be clear, conspicuous, and placed at the point of the claim — not buried in a landing-page footer.

What happens if I bid below the platform's effective floor? Bids under roughly $3 CPC frequently fail to win any impressions at all, regardless of how well-targeted the campaign is. Underbidding isn't a cost-saving strategy on this channel — it's a delivery-blocking one.

Is ChatGPT Ads pricing still based on the $60 CPM it launched at? No. That was the enterprise pilot launch price in February 2026. CPMs have since dropped to a typical $25–$60 range, and CPC bidding, running roughly $3–$18 depending on category, is now the platform's primary performance objective.

Do FINRA and CFPB rules really apply to ChatGPT ad creative, even if it's AI-generated? Yes. FINRA's guidance treats AI-generated content used in external communications as a communication subject to normal supervision and retention rules, and CFPB interpretive guidance treats AI-targeted advertising as within UDAAP authority — being AI-generated doesn't create an exemption.